This directory contains the API codebase. It's written in Go and is fully cross-platform (does not depend on any platform-specific libraries or C code).

The codebase follows idiomatic Go guidelines. Interfaces are preferred as a means of sharing functionality in order to facilitate easier mocking and testing. cmd/server/main.go contains the main setup code in a function named Initialise. The app itself starts up and blocks when App.Start is called. This fires up each of the components in the app in a goroutine then blocks until the root context is cancelled.

The library of choice for HTTP routing etc is go-chi. In src/api/ there is a directory for each API "Module" and the names of these must correspond to its respective API route except legacy because of how SA-MP legacy queries work.

Each handler goes in its own file, prefixed with h_ followed by a simple description or HTTP method of what it does. The setup code for each API module lives in a file named api.go.

For example:

api/
├module1/
│ ├api.go
│ ├h_get.go
│ └h_post.go
└module2/
  ├api.go
  ├h_get.go
  ├h_post.go
  └h_search.go

The naming for these files is not super strict, most of the time a HTTP method such as "get" or "post" should be fine, but other simple verbs are fine too. What matters is consistency and ease of navigation.

The database code is managed by Prisma. The schema is declared in the prisma directory in the root.

The workflow to make changes to the data model is:

1. Make the change
2. Generate new Go code
3. Generate migrations
4. Apply migrations to dev instance

Once a pull request is approved, if it has data model changes, an admin will apply those changes to the live DB.

SA-MP server queries are done by the Worker periodically querying the database for servers that haven't been updated within some time window. These servers are then passed to a Scraper which uses a Queryer to fetch updated information about the servers. The Worker then updates the database with the new data using the server.Repository

The docs/wiki part of the site (open.mp/docs) is very big. It contains around 3000 pages at the time of writing and will likely grow as more contributors translate the original pages. Because of this, Next.js and Vercel pre-rendering was taking upwards of 15 minutes for each pull request. Because we wanted faster build times, we sacrificed a bit of runtime performance for build-time. Page load times were only affected a tiny bit so it seemed a worthwhile tradeoff.

So, the solution to this was to move the static docs files to the API. Because the Docker image of this codebase just does ADD . ., the entire repo is available to the server. The docs/ directory is served as a http.FileServer at /docs/*.

There's also a Bleve search index built at image build time. This is done with indexbuilder which is a simple Go "script" that runs the Bleve index build code on the docs/ directory. This search index is then accessible via /docs/search?q=sometext. This is done at image build time because it only needs to be done once for any change to the files. It takes a few minutes usually.

This service uses OAuth2 for authentication. The endpoints are in src/api/auth/ and the implementation for each is in src/authentication. There are traces of code for password-based auth but this isn't exposed currently as OAuth2 is just safer, easier and results in reduced security responsibility on our part.